Dall-e-3 API - generation ID

Hi, I am really confused with all that stuff around seeds and generation IDs. Recently, seeds could be used to generate similar styled images. Now, seeds are not possible to use, if I understand right, but you can somehow use generation ID. At least you can use it within chatGPT. But in API reference, there is no mention about generation ID. On api generation image call, there is only described response as list of urls, and a timestamp. But no mention regarding generation ID. Do I understand it right, that generation ID is feature of chatGPT, or it is possible to retrieve generation ID also from API, and refer it in next api call?.. Url, metadata…?

2 Likes

Welcome to the forum.

At this time, there is no generation_id used in ChatGPT or the API. I think ChatGPT might hallucinate the concept sometimes and go along with it.

I’ve heard the DALLE team is working on this for future versions, though. For now, the best way is to “force” the prompt so that it doesn’t change and use the exact same language except a few words to get the same style/subject…

Hope that helps. We’ve got some great dalle3 threads, so hope you stick around!

2 Likes

am I correct that this feature is not currently (April 2024) hallucinated by chatgpt? This thread comes up on top of the search results.

Are we getting it in the API soon?

1 Like

The version of context from the DALL-E GPT (that has the capability to generate two images, while your GPT is denied this functioning parameter), showing the use of a previous image reference.

## dalle

// ... (description with instructions)
namespace dalle {

// Create images from a text-only prompt.
type text2im = (_: {
// The size of the requested image. Use 1024x1024 (square) as the default, 1792x1024 if the user requests a wide image, and 1024x1792 for full-body portraits. Always include this parameter in the request.
size?: "1792x1024" | "1024x1024" | "1024x1792",
// The number of images to generate. If the user does not specify a number, generate 2 images.
n?: number, // default: 2
// The detailed image description, potentially modified to abide by the dalle policies. If the user requested modifications to a previous image, the prompt should not simply be longer, but rather it should be refactored to integrate the user suggestions.
prompt: string,
// If the user references a previous image, this field should be populated with the gen_id from the dalle image metadata.
referenced_image_ids?: string[],
}) => any;

} // namespace dalle

This cannot be employed directly by stateless API, as it is an abstraction of a seed state that is maintained for that chat session alone. The actual random seed setting of a signed float32 initial state OpenAI has so far denied to API users for unknown reasons (like ensuring that developers never have feature parity with what OpenAI is selling in ChatGPT).

2 Likes